unity8-autopilot device unlock seems racy

Any ideas on this? We're seeing it on some of the MP testing jobs too it seems.

Oh yeah, it's a relatively simple fix, just need to retry the unlock a couple times. I've just been focused on other things. Is this causing a lot of pain?

As mentioned, it's not every job, but when it happens it brings things to a
halt pretty quickly. It can be pretty annoying for someone trying to land
something too I suspect if they need to retry, but spend time trying to
sort out if this is something they caused.

On Fri, Sep 26, 2014 at 7:23 AM, Michael Terry <email address hidden>
wrote:

> Oh yeah, it's a relatively simple fix, just need to retry the unlock a
> couple times. I've just been focused on other things. Is this causing
> a lot of pain?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1370644
>
> Title:
> unity8-autopilot device unlock seems racy
>
> Status in “unity8” package in Ubuntu:
> New
>
> Bug description:
> In some test runs, we see unlocking fail with:
> I: Unlock failed, script output: 'Error:
> GDBus.Error:org.freedesktop.DBus.Error.ServiceUnknown: The name
> com.canonical.UnityGreeter was not provided by any .service files
>
> It's not reliably reproducible, but probably happens at least once a
> day for us in CI
>
> To manage notifications about this bug go to:
>
> https://bugs.launchpad.net/ubuntu/+source/unity8/+bug/1370644/+subscriptions
>

I mentioned this on IRC, but for posterity: please try http://paste.ubuntu.com/8433042/ and let me know how that works for you

This bug was fixed in the package unity8 - 8.00+14.10.20141006-0ubuntu1

---------------
unity8 (8.00+14.10.20141006-0ubuntu1) utopic; urgency=low

  [ Andrea Cimitan ]
  * Tweak card header to match the spec
  * Add preview image slideshow (LP: #1351537)

  [ Michał Sawicz ]
  * Cache more things in memory, so flicking scopes should be faster
    (LP: #1336724)
  * Tweak card header to match the spec
  * Save texture memory by limiting sourceSize (LP: #1338430)

  [ Ying-Chun Liu ]
  * Add attributes to Preview. (LP: #1282460)

  [ Albert Astals ]
  * Update pot
  * Cache more things in memory, so flicking scopes should be faster
    (LP: #1336724)
  * Save texture memory by limiting sourceSize (LP: #1338430)
  * Clip the settings list
  * Fix unlocking from the left again
  * Add wait_ makes tests more reliable

  [ Michael Zanetti ]
  * fix fading out the launcher instead of sliding it out on left-edge
    minimizing an app.
  * Make the DashCommunicator async and more flexible to handle a
    lifecycle-suspended dash (LP: #1339883)

  [ Michael Terry ]
  * Retry unlock-device script if it fails, as there is always a risk of
    a small race with boot-up. (LP: #1370644)
  * Add pull-to-refresh functionality to scopes. (LP: #1368336)

  [ CI bot ]
  * Resync trunk
 -- Ubuntu daily release <email address hidden> Mon, 06 Oct 2014 08:03:23 +0000

Re-opening. There's still a race here. There's a chance that our dbus server will respond before the UI is actually ready to handle the incoming signal. Looking at it.

You can reproduce this latest race with the following script:

adb reboot && adb wait-for-device && adb shell 'while ! gdbus call --session --dest com.canonical.UnityGreeter --object-path / --method com.canonical.UnityGreeter.HideGreeter 2>/dev/null 1>&2; do echo -n .; done' && echo -e '\nDid it work?'

This exact race does not seem to have occurred in 8.00+14.10.20140930.2-0ubuntu1. But does in 8.00+14.10.20141006-0ubuntu1.

I don't think the appearance of this race is due to my previous unlock-device changes, as this script does not exercise the unlock-device command. This is something new that needs investigation.

OK, I have a possible fix, but need to clean it up and make it generic for a merge proposal. But it seems to work locally.

I plan to propose a merge tomorrow.